Professional Documents
Culture Documents
. I declare that this essay is work done as a part of the Part III Exami-
nation at the University of Cambridge under supervision of Professor
Coates. The essay states the primary aim as the generalization of the
Birch Lemma of Birch and Heegner about quadratic twists with prime
discriminants and those having any prescribed number of prime factors
but the majority of the high level mathematics involved in this lemma
required background knowledge more than taught or acquired at this
stage. The actual aim of essay was to understand the Conjecture of
Birch and Swinnerton-Dyer and the basic background related to the
problem. This essay is the result of my own work, and except where
explicitly stated otherwise, only includes material undertaken since the
publication of the list of essay titles, and includes nothing which was
performed in collaboration. The essay required explaining, for elliptic
curves defined over Q, the precise statement of the conjecture of Birch
and Swinnerton-Dyer, both in its weak and strong forms followed by
a discussion, without proofs, of the theorem of Kolyvagin-Gross Zagier
and the parity theorem of T. and V. Dokchister, and also some illu-
minating numerical examples in support of the conjecture. Then the
essay required to give an account, with detailed proof with explanation
where possible, of one of the results in the reference Quadratic twists
of elliptic curves (J. Coates, Y. Li, Y. Tian, S. Zhuai). The second ref-
erence Lectures on the Birch-Swinnerton-Dyer conjecture (J. Coates)
gives some background material for the first reference. Although not
up to date with the subsequent literature, Tates short article The arith-
metic of elliptic curves, as the third reference, was still a beautifully
simple and precise account of the conjecture.
1
THE CONJECTURE OF BIRCH AND
SWINNERTON-DYER
C(f¯) = {f¯(x, y, z) = 0} ⊂ P2
Definition 1.2 (Singular Point). We say that P = (a : b : c) is a
singular point on C if all partial derivatives at P vanish.
If a curve C has no singular points, it’s called a non-singular curve.
A cubic can have at most one singular point. For if it has more than
one singular point, then the line passing through those two points on
cubic will have multiplicity 4 which is not possible.
Elliptic curves.
Theorem 1.3 (Falting). If C is a curve of genus g > 1 over Q. Then
C(Q) is finite.
For g = 0, either C(Q) = ∅ or C ' P1 . When g ≥ 2, Faltings
theorem tells us that #C(Q) < ∞. Our concern is the remaining case
g = 1. If C(Q) is a genus one curve and C(Q) = ∅, there isnt much
to do. Assume that C(Q) contains at least one point, say O. Such a
curve is called an elliptic curve, and usually we denote C by E for an
elliptic curve.
Definition 1.4 (Elliptic Curve). An elliptic curve over a field K is
a non-singular projective algebraic curve E of genus 1 over K with a
chosen base point O ∈ E.
Group Law. Let P, Q, R ∈ E(K) where K is a field and let O be the
base point. There exists a binary operation ⊕ on E(K) such that:
(1) P ⊕ Q = Q ⊕ P
(2) P ⊕ O = P
(3) If a line L meets E at points P, Q, R, then (P ⊕ Q) ⊕ R = O
(4) Given P ∈ E(K), there exists R ∈ E(K) such that P ⊕ R = O
(Then we write R = P )
(5) (P ⊕ Q) ⊕ R = P ⊕ (Q ⊕ R)
Therefore, (E(K), ⊕) is an Abelian Group.
E : y 2 = x3 + 11x + 17
We will look for points on E in the field F5 . We first consider the table
of square in F5
y y2
0 0
1 1
2 4
3 4
4 1
Next we consider the cubes and the right hand side of our defining
equation for E in F5
x x2 x3 + x + 2
0 0 2
1 1 4
2 3 2
3 2 2
4 4 0
Equating the left and right side of the equation of the elliptic curve, we
find there are four points in E(F5 ), and they are (1,2), (1,3), (4,0) and
O.
Thus, counting points on an elliptic curve is simple and relatively quick
in a finite field compared to the rationals.
y 2 + a1 xy + a3 y = x3 + a2 x2 + a4 x + a6 .
x y
Since replacing (x,y) by ( , ) causes ai to become ui ai , if we choose
u2 u3
u divisible by a large power of π (where π represents a uniformizer of
the ring of integers-here Z), then we find a Weierstrass equation with
all coefficients ai ∈ Z such that |∆| is as small as possible. This is
called the minimal Weierstrass equation.
Let Ẽ represent the reduced curve for the minimal Weierstrass equa-
tion
The idea of Birch and Swinnerton Dyer was that the larger E(Q) is,
the larger the E(Fp )s should be on average as p varies. But how can
one measure the average size of the E(Fp )s?
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 7
near a line with a certain slope. And that slope was the rank (r) of
E as defined. Their jagged graphs led Birch and Swinnerton Dyer to
guess that if E is an elliptic curve over Q of rank r, then
πE (x) ∼ C(log x)r
for some non-zero, positive constant C depending only on E.
The function πE does not behave very nicely and therefore is difficult
to work with. Birch and Swinnerton-Dyer were then led to a modified
conjecture, using the L-function of E in place of πE .
Note: The rank we have been talking so far is the algebraic rank
and will be denoted by r.
3
converges for <(s) > .
2
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 9
The terms due to singularity for which p|∆ is known as the fudge
factor. We want to show that L(s, E) continues analytically to be
analytic function in the entire s-plane. This requires knowing:
(1) Every elliptic curve E is modular. (Wiles-1995, Taylor and
Wiles-1995 and Breuil et al.-2001)
(2) L-function of modular form has analytic continuation and a
functional equation. (Hecke)
A functional equation should relate the value of L(E, s) to the value of
the same function evaluated at some point other than s. For this we
define a new function Λ. We define
p
( C(E))s
Λ(E, s) = Γ(s)L(E, s)
(2π)s
where C(E) is the conductor of E.1 and Γ(s) is the Gamma function
which is analytic for all values except for s ∈ Z≤0 . So we have formed
a new function out of L(E, s) by multiplying with gamma function and
throwing in some other exponential factors.
Modular Forms.
Definition 2.5 (Modular Forms). Let T0 (C(E)) be a subgroup of SL(2, Z)
such that T0 (C(E)) =
a b
{ | a, b, c, d ∈ Z and ad − bc = 1, c ≡ 0 mod C(E)}
c d
Let H be the upper half-plane H = {z ∈ C, =(z) > 0} and let τ ∈ H
and q = e2πiτ . Then define the Modular form associated with the
Dirichlet series of an elliptic curve as follows
∞
X
fE (τ ) = an q n
n=1
We can guess to replace the right hand side with partial products, up
to p ≤ x for any real number x greater than the largest prime in S.
When we do, we get
(1 − ap p−1 )−1 (1 − ap p−1 + p−1 )−1
Q Q
RHS up to x =
p|∆,p≤x p6 |∆,p≤x
Taking reciprocals,
(1 − ap p−1 ) (1 − ap p−1 + p−1 )
Q Q
1/RHS up to x =
p|∆,p≤x p6 |∆,p≤x
Np p − ap
Now let us look at . For p |∆, this quotient is = 1 − ap p−1 .
p p
For p6 |∆, it is (p − ap + 1)/p = 1 − ap p−1 + p−1 . In every case, it is the
corresponding factor in the product we just wrote down. So we can
rewrite this product as
Q Np
1/RHS up to x =
p≤x p
We recognize the right hand side above as πE (x). Thus we get,
1
RHS up to x =
πE (x)
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 11
The existing methods for determining the rank r and basis points
of E(Q) all depend more or less on the Birch and Swinnerton Dyer
Conjecture and rely on the basic exact sequence above. Since, the
Selmer group can be computed depending on m,
r = rank(E(Q)/mE(Q))
is all a matter of computing the Tate-Shafarevich Group.
Let III(E | Q)div be the maximal divisible subgroup of III(E | Q).
Theorem 2.9 (Cassels-Tate). There is a canonical alternating bilinear
form on III(E | Q)/III(E | Q)div .
Corollary 2.10. The vector space (III(E | Q)/III(E | Q)div )[p] has
even Fp -dimension.
Corollary 2.11. If III(E | Q)div [p] = 0, then # III(E | Q)[p] is a
square.
3. Weak Parity Theorem of T. and V. Dokchitser
Classical Galois Cohomology shows that, for some integer tE,p ≥ 0,
we have:
Qp
III(E | Q)[p]=( )tE,p
L
(a finite group)
Zp
Qp
so that III(E | Q)div [p]=( )tE,p .
Zp
Conjecture: The group III(E | Q) is finite.
The answer to the above conjecture is undoubtedly one of the major
problems in Number Theory. However for elliptic curves with rank
greater than 1, it hasn’t been shown. If it happens to be true, than
tE,p = 0 for every p. So, the factor tE,p = 0 is also important to us and
the only main fact known regarding it was given by T. & V. Dokchitser.
Theorem 3.1 (T. & V. Dokchitser). Weak Parity Theorem. Let E(Q)
be any elliptic curve and p be a any prime. Then we have:
ρ ≡ tE,p + r mod 2.
So, if Tate-Shafarevich group was proven to be finite i.e. tE,p = 0,
then the weak parity theorem would imply the strong parity conjecture
ρ ≡ r mod 2. The main importance of the parity conjecture and
finiteness of the Tate Shafarevich group is that they can together imply
the Birch and Swinnerton Dyer conjecture. And we know finiteness of
Tate Shafarevich group alone with Weak Parity Theorem would imply
the Parity conjecture. Ultimately, all we care about is understanding
the mysterious Tate-Shafarevich Group.
14 THE CONJECTURE OF BIRCH AND SWINNERTON-DYER
The problem with this is that the trace might only give points of finite
order. In 1983, Gross and Zagier proved the following theorem.
Theorem 5.3 (Gross and Zagier). Assume that E is modular. If
L(E, 1) = 0 and L0 (E, 1) 6= 0, then Heegner’s construction gives a
point of infinite order on E over rationals.
Such points are called Heegner points. A Heegner point8 is a point on
a modular curve that is the image of a quadratic imaginary point of the
upper half-plane. This is a remarkable result. It means that one can
prove the existence of infinitely many rational solutions to an equation
by testing the vanishing of a certain analytic function. The GrossZagier
theorem (Gross & Zagier 1986) describes the height of Heegner points
in terms of a derivative of the L-function of the elliptic curve at the
point s = 1. In particular if the elliptic curve has (analytic-ρ) rank 1,
then the Heegner points can be used to construct a rational point on
the curve of infinite order (so the Mordell Weil group has rank (r) at
least 1). More generally, Gross, Kohnen & Zagier (1987) showed that
Heegner points could be used to construct rational points on the curve
for each positive integer n, and the heights of these points were the
coefficients of a modular form of weight 3/2.
The second assertion uses Kolyvagins method of Euler systems and
an infinite family of Heegner points. In 1988, Kolyvagin built on this
work of Gross and Zagier and gave the following theorem.
Theorem 5.4 (Kolyvagin). Assume E is modular. Then:
(1) If L(E, 1) 6= 0, then #E(Q) is finite.
(2) If L(E, 1) = 0 but L0 (E, 1) 6= 0, then E(Q) = Z ⊕ finite group.
Kolyvagin’s method involved introducing some new algebraic tech-
niques, in particular the notion of an Euler System. It also involved
some analytic results on the existence of twists of the L-functions of E
which have no zero or a zero to the first order.9 Let us start with finite
dimensional p-adic representation T of the Galois group of a number
field K. We then define an Euler system for T to be a collection of coho-
mology classes (which Kolyvagin calls derivative classes) cF ∈ H 1 (F, T )
for a family of abelian extensions F of K, with a relation between cF 0
and cF whenever F ⊂ F 0 . Kolyvagins machinery is designed to use
these derivative classes to provide an upper bound for the size of the
Selmer group associated to the Cartier dual10 T ∗ . One common feature
8They were defined by Bryan Birch and named after Kurt Heegner.
9See Quadratic Twists of Elliptic Curves.
10Given any finite flat commutative group scheme G over S, its Cartier dual is
the group of characters, defined as the functor that takes any S-scheme T to the
18 THE CONJECTURE OF BIRCH AND SWINNERTON-DYER
of all Euler systems is that they are closely related to special values
of L-functions. An important benefit of this connection is that the
bounds on Selmer groups that come out of Euler system machinery are
then linked to L-values (leading term of L-functions).
Combining these two statements with the basic exact sequence:
0 → E(Q)/mE(Q) → S m (E | Q) → III(E | Q) → 0
proves that if ρ ≤ 1, then r = ρ and #III(E| Q) ≤ C.
The theorem of Kolyvagin-Gross Zagier also answers the rationality
question for ρ ≤ 1.
LE
Theorem 5.5. If ρ ≤ 1, then ∈ Q.
c∞ (E)R∞ (E)
Case when ρ=0. From the work of Kolyvagin, we get both E(Q) and
III(E| Q) are finite. It follows that LE = L(E, 1) and R∞ =1. We thus,
define
L(E, 1)
Lalg (E, 1) =
c∞ (E)
which is a rational number. Due to limited knowledge of the exact
Birch and Swinnerton Dyer formula (even in above case), we break the
exact formula up into a p-part for all primes p. Iwasawa theory gives
considerable knowledge about this p-part.
Conjecture: Assuming ρ = 0, we have for all primes p,
ordp (Lalg (E, 1)) = ordp (Tam(E))−2ordp (#(E(Q)))+ordp (#(III(E | Q)))
Q
where Tam(E) = cq (E).
q|C(E)
Theorem 5.6 (Rubin). Assume that L(E, 1) 6= 0 and that E has com-
plex multiplication11.Then p-part of the Birch and Swinnerton Dyer
Conjecture holds for√all primes p 6= 2. (In addition, if E has complex
multiplication by Q −3, we must exclude p = 3 as well as p = 2).
When E does not have complex multiplication, we have the following
weaker theorem.
Theorem 5.7 (Kato, Skinner-Urban). Assume that L(E, 1) 6= 0. Then
the p-part of the Birch and Swinnerton Dyer holds for all good ordinary
primes p except those in some specified list, which certainly includes
p = 2.
abelian group of group scheme homomorphisms from the base change GT to Gm,T
and any map of S-schemes to the canonical map of character groups.
11If End(E) is strictly larger than Z, then we say E has complex multiplication.
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 19
and hence ρ > 1. Now computing L00 (E, 1) approximately will give us
ρ = 3 if it is. So if ρ < 1, then we can find the exact value of ρ by using
it’s root number for determining parity, modular forms to establish if
ρ = 0 , the theorem of Kolyvagin-Gross-Zagier to distinguish ρ = 1
from ρ = 3 and numerical evaluation of Lj (E, 1). If ρ > 3, there is no
way of determining ρ rigorously. For example if ρ = 4, we can guess
the parity and compute L00 (E, 1) approximately close to zero, but we
can’t show it is exactly zero. Similarly for ρ = 5, we can show it’s odd
and compute L000 (E, 1) approximately close to zero but there is no way
of showing that it is exactly zero. There are 614308 isogeny classes
of elliptic curves with conductor C(E)≤140000. All have ρ ≤3 and in
every case r = ρ.[7]
Range of C(E) # r=0 r=1 r=1 r= 3
0-9999 38042 16450 19622 1969 1
10000-19999 43175 17101 22576 3490 8
20000-29999 44141 17329 22601 4183 28
30000-39999 44324 16980 22789 4517 38
40000-49999 44519 16912 22826 4727 54
50000-59999 44301 16728 22400 5126 47
60000-69999 44361 16568 22558 5147 88
70000-79999 44449 16717 22247 5400 85
80000-89999 44861 17052 22341 5369 99
90000-99999 45053 16923 22749 5568 83
100000-109999 44274 16599 22165 5369 141
110000-119999 44071 16307 22173 5453 138
120000-129999 44655 16288 22621 5648 98
130000-139999 44082 16025 22201 5738 118
0-139999 614308 233979 311599 67704 1026
We will give some illuminating numerical examples in support of the
conjecture:[7]
• r=0:12 E: 11a1 has coefficients [0,-1,1,-10,-20] and conductor
L(E, 1) 1
11. The modular symbols give us = exactly. Thus,
Q Ω(E) 5
ρ = 0 and ρ = r. Moreover, cp = c11 = 5 and #E(Q)tors = 5.
p
Then strong form of Birch and Swinnerton Dyer Conjecture
predicts that:13:
12R = 1 when ρ=0
∞
13This
can be verified by careful application of known results: see R. L. Miller,
Proving the Birch and Swinnerton-Dyer conjecture for specific elliptic curves of
analytic rank zero and one, arXiv:1010.2431v2 [math.NT]
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 21
6. An Interesting Example
Congruent Number Problem. [1] A concrete example of the num-
ber theoretic problem that can be solved if Birch and Swinnerton Dyer
Conjecture is proven to be true is the Congruent Number Problem.
This was shown by mathematician Jerrold Tunnell (1950-). Let n ∈ Z.
Then can we find a right triangle with sides a, b and c all rational
numbers with area n? In other words, we need to solve simultaneous
equations:
a2 + b 2 = c 2
ab
=n
2
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 23
15Usually
Isom(C)=Aut(C). But for elliptic curves, Aut(E) is defined as all
isomorphisms from E to E that take O to O. Thus for elliptic curves Aut(E)6=
Isom(E)
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 25
Let E(K) be an elliptic curve and let us ignore the base point O,
then the twists of E(K) correspond to the set H 1 (GK̄/K ,Isom(E)).
Now Isom(E) has two obvious subgroups namely Aut(E) and E (E is
identified as the set of translations {τp } in Isom(E)). Generally the
twist of E corresponds to the set H 1 (GK̄/K , Aut(E)). But such a twist
should be called a twist of a pair (E, O).
Theorem 6.6. [5] Let E(K) be an elliptic curve.
• The natural inclusion Aut(E)⊂Isom(E) induces an inclusion:
H 1 (GK̄/K , Aut(E)) ⊂ H 1 (GK̄/K , Isom(E))
We identify the latter set by Twist(E(K)) and the former by
Twist((E,O)(K)).
• Let C(K) ∈ Twist((E,O)(K)). Then C(K)6= φ, and so C(K)
can be given the structure of an elliptic curve over K. (C is not
generally K-isomorphic to E).
• Conversely, if E0 (K) is an elliptic curve which is isomorphic to
E over K̄, then E0 (K) represents an element of Twist((E,O)(K)).
If the characteristic of K is not 2 or 3, then the elements of Twist((E,O)(K))
can be described explicitly by the following theorem.
Theorem 6.7. [5] Assume that char(K) 6= 2, 3, and let
2 if jE 6= 0, 1728
n = 4 if jE = 1728
6 if j = 0
E
16A
complex number a is a simple zero of f , or a zero of multiplicity 1 of f , if
f can be written as f (z) = (z − a)g(z), where g is a holomorphic function g such
that g(a) is not zero.
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 27
Then, if k is any integer ≥1, there are infinitely many square integers
M, having exactly k prime factors, such that L(E (M ) , s) has a zero at
s=1 of order 1. Similarly, if k is any integer ≥2, there are infinitely
many square-free integers M, having exactly k prime factors such that
L(E (M ) , s) does not vanish at s=1.
By applying the theorem of Gross and Zagier to the point yK and from
the properties of canonical height, we get yK has infinite order if and
only if L0 (E/K, 1) 6= 1. Further we note the following proposition:
Let K be a quadratic number field with discriminant d, (here −l0 ),
and E be an elliptic curve over Q. Then:
(1) rank E(K)= rank E(Q)+ rank E (d) (Q) and
(2) L(E/K) = L(E | Q, s)L(E (d) | Q)
The following corollary[3] immediately follows which implies the asser-
tion of Theorem we want to prove when k = 1.
Corollary 7.2. [3] Let E be an elliptic curve defined over Q with con-
ductor C. Let f : X0 (C) → E be a fixed modular √ parametrisation
and assume that f ([0]) ∈ / 2E(Q). Let K = Q −l0 as defined above
and let the Heegner hypothesis hold. Let C be an ideal in O such that
O/C ∼= Z/CZ and let P = P1 as defined above be the corresponding
Heegner Point of conductor 1. Then the complex L-function L(E/K, s)
of E over K has a simple zero at s = 1, L(E, s) does not vanish at
s = 1 and L(E (−l0 ) ,s) has a simple zero at s = 1.
Proof. From previous theorem, we get that yK is of infinite order. So,
we have constructed a point of infinite order and so by the theorem of
Gross and Zagier, we get that the complex-L series of E over K has a
simple zero at s = 1. Then, as L(E/K, s) = L(E, s)L(E −l0 , s). Also
if this limit exists. If the natural density exits, then it is actually equal
to the analytic density
25This really depends on the choice of q. We define frobenius symbol of p in
L/K to be the conjugacy class {F robq : q | p}. Note that in the case of an abelian
group, this set contains only a single element- though formally the element and the
set containing this element are two different objects.
34 THE CONJECTURE OF BIRCH AND SWINNERTON-DYER
P 1
s
p∈S #(OK /p)
dan (S) = lim+1 P
s→1 1
s
p prime #(OK /p)
The converse is however not true, however, there are cases where the
analytic density exists but the natural does not. However, the Cheb-
otarev Density theorem is valid with either notion of density.
Tate Module. Given an abelian group A and a prime number p, the
p-adic Tate module of A is
Tp (A) = lim A[pn ]
←
n n
where A[p ] is the p torsion of A (i.e. the kernel of the multiplica-
tion by pn map and the inverse limit is over positive integers n with
transition morphism given by the multiplication by p map
A[pn+1 ] → A[pn ]
Thus, the Tate module encodes all the p-power torsion of A. It is
equipped with the structure of Zp -module via z(an )n = ((z mod pn )an )n .
Lemma 7.3. P [3]n Let E be an elliptic curve over Q with a primitive
cusp φ = an q . Assume that E possesses a sensitive supersingular
n P
prime q1 , which is inert in K. For each integer r≥ 2, define r to be
the set of all primes q 6= q1 such that:
• q ≡ 1 mod 4,
• aq ≡ 0 mod 2r ,
• (q, C) = 1 and C is a square modulo q, and
• q is inert in K.
P
Then r is infinite of positive density in the set of primes.
√
Proof. Let J=Q( C, E[2r ]), and note √ that K ∩ J = Q. This is because
l0 is totally ramified in K/Q as K=Q −l0 and only those primes di-
viding 2C can ramify in J. As we know l0 ≡ 3 mod 4. So (l0 ,2C)=1.
This means that the prime l0 is not ramified in J.
Now since C is a square modulo q1 , we get (q1 , C) = 1 and since q1 ≡ 1
mod 4, we get that q1 is odd. So (q1 ,2C)=1 (By definition of Sensitive
Supersingular prime). This means that even q1 is not ramified in J.
Consider JK. Then any prime dividing 2l0 q1 C is ramified in JK.
χD (σ)f (PR )σ
P
zD =
σ∈ Gal(HR /K)
Obviously, zR = yR .
Q
Lemma 7.7. [3] For all positive divisors D of R, define bD = aq ,
q|R/D
where the product is taken over all primes q dividing R/D. Then we
have
zD = bD yD
Let D be any positive proper divisor of R prime to C such that (q, CD) =
1 where q√is a prime in R − D and we know q is inert in K. Now we
know K( D) is contained in HD , then the assertion easily follows from
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 39
(7.15) = zD
Now we are ready to prove our assertions about the Heegner point
yR . It is convenient to assume that T = f ([0]) is of exact order 2.
This can always be done by composing f with multiplication by an
odd integer on E, and we shall assume for the rest of the proof that
we have done this. Define
f (PR )σ
P
ψR =
σ∈ Gal(HR /HR )
ψR + σ(ψR ) = 0
Adding the above two equations, we get,
(7.16) yR = 2ψR
(7.17) ⇒yR + yR = 2(ψR + ψ R )
(7.18) ⇒yR + yR = 2T = 0
(7.19) ⇒yR = yR
as T is of order 2. Now since yR = 2ψR , we note that yR + σ(yR ) = 0.
This means that
σ(yR ) = y R .
This means that the point yR is mapped to its negative by √the non
√
trivial element of the Galois group of K( q1 )/K in E(Q( −l0 R).
√
Thus √ yR ∈ E(Q( −l0 R)− which is the subgroup those points of
E(Q( −l0 R) which are mapped to their negative. Now √ suppose we can
write a heegner point yR = 2w + t where w ∈ E(Q( −l0 R)− and t is a
torsion point. This implies that ψR = w + t0 where t0 lies in E(HR )[2∞ ].
But we know from the first lemma that E(HR√)[2∞ ] = E(Q)[2]. Thus
we get t0 ∈ E(Q)[2]. This means ψR ∈ E(Q( −l0 R)− which implies
ψR + ψ R = 0
which means that√T = 0 and contradicts
√ the fact that order of
√ T is 2.
Thus √yR ∈ E(Q( −l0 R)− +E(Q( −l0 R)tor and yR ∈ / 2E(Q( −l0 R)− +
E(Q( −l0 R)tor . This proves our assertion for r=1. Now suppose that
r > 1. Let y1 = T rHR /HR f (PR ) and for any positive divisor D of R
divisible by q1 , let zD = T rHR /K(√D) f (PR ). Then the points zD and y1
are related by the following relation
zD = 2r y1 ∀ r > 1
P
q1 |D|R
It immediately follows from the above formula and our case r = 1,
that:
zD = 2r ψR
P
yR +
D|R,D6=R
because yR = zR as seen before. If D 6= R, we can write bD = 2r eD for
some integer eD . This follows from the fact that aq ≡ 0 mod 2r . Now
as zD = bD yD by the lemma above, we get:
yR = 2r uR with uR = ψR −
P
eD yD
D|R,D6=R
Also we know bD = 0 if q1 does not √divide D which√ means eD = 0. It
follows that the class of yR in E(K( R))/2r E(K R) maps to zero in
E(HR )/2r E(HR ). But we have the inflation restriction exact sequence:
THE CONJECTURE OF BIRCH AND SWINNERTON-DYER 41
√ √
0 → H 1 (Gal(HR /K( R)), E(HR )[2r ]) → H 1 (K( R), E[2r ]) →
H 1 (HR , E[2r ])
and the kernel of the left of this sequence is killed by 2, because,
as mentioned
√ before, E(HR )[2∞ ] = √E(Q[2]). It follows that 2yR ∈
r
2 E(K( R)) for some y ∈ E(K( R)) and t ∈ E(Q)[2]. Then
2r uR = 2r−1 y + t. This gives y = 2ur√+ s for some s ∈ E(Q)[2]. We
want to show that y belongs to E(Q( −l0 R))− .
√ √ √
Let σ ∈ Gal(HR /K) that maps q1 to − q1 , and fixes all qi for
2 ≤ i ≤ r. We claim that:
σ(y) + y = 0
Now ψR = T rHR /HR f (PR ) and
σ(ψR ) + ψR = T rHR /K(√q2 ,..,√qr ) f (PR ) = 0
as aq1 = 0. If eD 6= 0, then D is a positive
√ divisor of R that is divisible
by q1 . Thus the restriction of σ to K( D) must be a non-trivial element
of the Galois group of this field over K, and so
yD = vD − σ(vD )
where vD = T rHD /K(√D) f (PD ). But since q1 |D and aq1 = 0, we have
vD + σ(vD ) = 0
Adding the above two equations, we get:
(7.20) yD = 2vD
(7.21) ⇒σ(yD ) + yD = 0
P
Now y = 2uR + s and uR = ψR − eD yD , and from above we
D|R,D6=R
have:
σ(ψR ) + ψR = 0 and σ(yD ) + yD = 0
Thus, it follows that
σ(y) + y = 0
Next we claim that y +y = 0. Consider all those divisors D of R, which
are not equal to R and for which eD 6= 0. We assert that:
yD + y D = 0
√
For those D just mentioned, K( D) is a subfield of HD which is a
subfield of R and since (wC P(HR ) = τ P(HR )), we have:
ψD + ψ D = T
42 THE CONJECTURE OF BIRCH AND SWINNERTON-DYER
that vD +vD = 0 Now ψD = T rHD /HD f (PD ) and vD = T rHD /K(√D) f (PD )
√
but K( D) is a subfield of HD . This implies that
(7.22) vD + vD = T
(7.23) ⇒ 2(vD + vD ) = 0
(7.24) ⇒ yD + yD = 0
(7.25) ⇒ uR + uR = ψR + ψ R
(7.26) ⇒y+y =0
(7.27) ⇒ σ(y) = y
p
(7.28) ⇒ y ∈ E(Q( −l0 R))−
p
(7.29) ⇒ yR ∈ 2r−1 E(Q( −l0 R))− + E(Q)[2]
√
Now √suppose yR = 2r y 0 + t with y 0 ∈ E(Q( −l0 R))− and t ∈
E(Q( −l0 R))tor . √Let m denote an odd integer that annihilates the
odd part of E(Q( −l0 R))tor , we then have
m(ψR − y 0 − eD yD ) ∈ E(HR )[2∞ ] = E(Q)[2]
P
D|R,D6= R
Corollary 7.8. [3] Under the same hypothesis as in the theorem proved
above, for all R = q1 ...qr with r ≥ 1 we have that:
• the complex L-series of E (R) does not vanish at s = 1, and both
E (R) (Q) and the Tate-Shafarevich group of E (R) are finite, and
• the complex L-series of E (−l0 R) has a simple zero at s = 1,
E (−l0 R) (Q) has rank 1, and the Tate-Shafarevich group of E (−l0 R)
is finite.
Proof. Since yR has infinite order, it follows from the Special Case of
Explicit Gross and Zagier formula applied current the hypothesis that
L(E/K, χR , s) must have a simple zero at s = 1. But as seen before:
References
[1] Avner Ash & Robert Gross, Elliptic Tales: Curves, Counting, and Number
Theory
[2] John Coates, Lectures on the Birch and Swinnerton-Dyer Conjecture, Em-
manuel College, University of Cambridge, and POSTECH
[3] John Coates, Yongxiong Li, Ye Tian and Shuai Zhai, Quadratic Twists of Ellip-
tic Curves, Proceedings of the London Mathematical Society Advance Access
published December 6, 2014.
[4] Karl Rubin and Alice Siverberg, Ranks of Elliptic Curves, Bulletin (New Series)
of The American mathematical Society, Volume 39, Number 4, Pages 455474-
July 8, 2002.
[5] Joseph H. Silverman, The Arithmetic of Elliptic Curves
[6] Susanne Schmitt and Horst G. Zimmer, Elliptic Curves- A Computational Ap-
proach
[7] John Cremona, Numerical evidence for the BirchSwinnerton-Dyer Conjecture,
University of Warwick at the BSD conference, Cambridge-4 May 2011.
[8] Li Cai, Jie Shu, and Ye Tian, Explicit Gross Zagier and Waldspurger Formula
[9] Kenneth Ireland, Michael Ira Rosen, A Classical Introduction to Modern Num-
ber Theory
[10] Hendrik Lenstra, The Chebotarev Density Theorem
[11] Adam Mickiewicz, Average Ranks of Elliptic Curves, Based on Mini-course by
Prof. Tim Dokchister, University in Poznan, 14 16.05.2014, Notes taken by
Jedrzej Garnek.
[12] Allison Arnold-Roksandich, Christopher Towse-Advisor, Michael E. Orrison-
Reader, There and Back Again: Elliptic Curves, Modular Forms, and L-
Functions, Department of Mathematics, Harvey Mudd College, May 2014.
[13] Lecture notes for the Part III course (2004-5) Elliptic Curves, University of
Warwick.
[14] Ye Tian, Congruent numbers and Heegner Points, arxiv:1210.8231v1
[math.NT]
[15] Francesca Gala, Prof. Jan Nekovar, Heegner points on X0 (N ), Universita Degli
Studi Di Milano , Facolta di Scienze Matematiche, Fisiche e Naturali ‘ Corso di
Laurea in Matematica.
[16] B.Gross, Heegner points on X0 (N ), Modular Forms (ed. R.A. Rankin; Ellis
Horwood, Chichester, 1984)
[17] John T. Tate, The Arithmetic of Elliptic Curves, Cambridge Mass, Inventiones
math, Vol. 23.